Picard MergeBam Alignment issue

[NitaC]

Hello everyone. I've scourged the internet and this forum in particular and I have not been able to get past this problem. Ultimately, I want to be able to use Picard's Mark Dups.

I have paired-end reads aligned to the human genome using GSnap. From previous experience, I already know that to use MarkDups I had to go through all of the pre-processing just to get the bam file up to Picard's standards. And so, I used FastQtoSam to make an unaligned bam file (queryname sorted). I then went to merge this unaligned file with my original GSnap output. I get this error:
Exception in thread "main" java.lang.IllegalStateException: Aligned record iterator (HWI-ST970:586:C2PY6ACXX:1:1115:10000:27368) is behind t
he unmapped reads (HWI-ST970:586:C2PY6ACXX:1:1115:10000:3435)
at net.sf.picard.sam.AbstractAlignmentMerger.mergeAlignment(AbstractAlignmentMerger.java:332)
at net.sf.picard.sam.SamAlignmentMerger.mergeAlignment(SamAlignmentMerger.java:156)
at net.sf.picard.sam.MergeBamAlignment.doWork(MergeBamAlignment.java:236)
at net.sf.picard.cmdline.CommandLineProgram.instanceMain(CommandLineProgram.java:177)
at net.sf.picard.sam.MergeBamAlignment.main(MergeBamAlignment.java:205)

I even tried merging my unaligned bam file with a sorted version of my aligned bam file (I tried both using coordinate sorted and queryname). In each try I get the same exact error. I am using the latest version of Picard and I can find nothing else on the internet that has solved my problem. Thanks in advance for any suggestions.

[NitaC]

I moved my files to a different server, used the same commands, with the same version of picard and now I am getting a different error.

It appears that the merge almost finishes but then something happens.

[Mon Nov 18 10:15:34 EST 2013] net.sf.picard.sam.MergeBamAlignment done. Elapsed time: 27.55 minutes.
Runtime.totalMemory()=5144641536
To get help, see http://picard.sourceforge.net/index.shtml#GettingHelp
Exception in thread "main" java.lang.NullPointerException
at net.sf.picard.sam.AbstractAlignmentMerger.updateCigarForTrimmedOrClippedBases(AbstractAlignmentMerger.java:529)
at net.sf.picard.sam.AbstractAlignmentMerger.transferAlignmentInfoToFragment(AbstractAlignmentMerger.java:418)
at net.sf.picard.sam.AbstractAlignmentMerger.transferAlignmentInfoToPairedRead(AbstractAlignmentMerger.java:432)
at net.sf.picard.sam.AbstractAlignmentMerger.mergeAlignment(AbstractAlignmentMerger.java:274)
at net.sf.picard.sam.SamAlignmentMerger.mergeAlignment(SamAlignmentMerger.java:156)
at net.sf.picard.sam.MergeBamAlignment.doWork(MergeBamAlignment.java:236)
at net.sf.picard.cmdline.CommandLineProgram.instanceMain(CommandLineProgram.java:177)
at net.sf.picard.sam.MergeBamAlignment.main(MergeBamAlignment.java:205)

Please help. I am out of ideas for this.

[EricHaugen]

At least one of the bam files may not be sorted in precisely the manner Picard expects. You could try removing that assumption with the Picard parameter "ASSUME_SORTED=false".

Or try "samtools sort" on each bam followed by a "samtools merge", instead of Picard.

Only the GSnap alignment, and not the unaligned reads, will be relevant when running Picard's MarkDuplicates anyway. Duplicates are calculated from aligned positions, not the content of the sequences.